DC HVAC common mistakes
DC HVAC · Understand the traps and how to avoid them.
Treating kW of IT load and kW of cooling capacity as directly interchangeable, so the candidate sizes or evaluates cooling against IT power without accounting for heat gain through envelope, lighting, UPS losses and fan heat.
Most study material quotes IT load in kW and cooling units in kW, and candidates shortcut to a 1:1 comparison. This works only for the IT heat itself; total room sensible load is always higher, so capacity questions and redundancy questions get answered against the wrong baseline.
What to do instead
Always separate IT load from total cooling load. Build the load as IT + UPS/PDU losses + lighting + envelope + fan heat, and state which figure a question is asking about before calculating. In efficiency questions, be explicit whether the ratio uses IT power or total facility power.
Assuming raised-floor tiles are the primary airflow path and that adding perforated tiles anywhere increases cooling, rather than treating the plenum as a pressure-managed volume where tile placement and open area must match the CRAC/CRAH output.
Raised-floor delivery is the classic data centre image, so candidates apply it to every scenario including slab-floor halls, in-row cooling and rear-door heat exchangers. They also forget that excess open tiles depressurise the plenum and starve distant racks.
What to do instead
Before answering airflow questions, identify the delivery architecture (perimeter raised floor, in-row, in-rack, overhead ducted, rear-door). For raised floor, check that total tile open area is balanced against unit airflow and that tiles sit only in cold aisles in front of loaded racks. Treat blanking panels and grommets as part of the same pressure budget.
Confusing containment types: describing cold-aisle containment as if it were hot-aisle containment, or claiming containment removes the need for blanking panels and cable sealing.
Both containment strategies look similar in diagrams and both are described as 'separating hot and cold air'. Candidates memorise the label rather than the physical consequence: cold-aisle containment pressurises the cold aisle and needs a sealed floor and doors, while hot-aisle containment captures hot exhaust and returns it to the unit.
What to do instead
Learn each by what it encloses and what it pressurises. Cold-aisle containment: cold air is trapped and pressurised, so floor sealing and door integrity matter most, and the operator works in hot aisles. Hot-aisle containment: exhaust is captured and ducted or returned, and the operator works in cold aisles. In both cases, blanking panels and cable cut-out seals remain necessary because containment does not fix rack-internal bypass.
Mixing up the metrics: computing PUE as IT power divided by total power, treating PUE as a measure of IT efficiency, or using DCIE and PUE interchangeably without inverting.
The formulas are simple ratios and easy to invert under time pressure. Candidates also over-interpret PUE as an efficiency score for servers, when it only describes overhead beyond IT load, and they forget that PUE varies with load and outdoor conditions.
What to do instead
Anchor on the definition: PUE = total facility power / IT power, always greater than or equal to 1. DCIE = 1 / PUE, expressed as a percentage. Sanity-check any answer: a PUE below 1 is impossible, and a PUE of 1.0 means zero overhead. Remember that PUE says nothing about how efficiently the IT equipment itself uses that power.
Ignoring the difference between sensible and total capacity, and between nominal and actual capacity at design conditions, when matching a unit to a load.
Unit datasheets quote nominal capacities at standard rating conditions, often with a total-to-sensible split. Candidates compare the nominal number directly to room load and conclude the unit is adequate, forgetting that the data centre load is almost entirely sensible and that capacity falls as entering air temperature rises or airflow changes.
What to do instead
For data centre work, size on sensible capacity at the actual entering air conditions and required airflow, not nominal total capacity. Check the sensible heat ratio and confirm the unit's sensible capacity at the design return temperature exceeds the room sensible load, then apply redundancy and derating on top.
Treating N+1 redundancy as sufficient protection without checking whether the redundant unit is on the same power path, the same chilled water loop, or the same controller, and without distinguishing concurrent maintainability from fault tolerance.
Redundancy is usually taught as a count (N+1, 2N) rather than a topology. Candidates answer 'N+1, so it is protected' and miss shared single points of failure such as a common chilled water header, a shared electrical panel or a common control system.
What to do instead
For each redundancy question, trace the full path: power source, cooling source, distribution and controls. Ask whether the spare capacity is available during maintenance of any single component (concurrent maintainability) and whether a single failure can take out both the primary and the spare. N+1 on a shared loop is not the same as N+1 on independent loops.
Assuming a lower supply air temperature always improves cooling, or that increasing airflow always lowers rack temperatures, without considering coil capacity, humidity and fan energy.
Candidates apply a simple 'colder is better' or 'more air is better' rule. In practice, lowering supply temperature can reduce chiller efficiency and may not address hot spots caused by bypass or recirculation, while raising airflow increases fan power and can disturb containment pressure balance.
What to do instead
Diagnose the hot spot first: is it insufficient airflow at the rack, recirculation from poor containment, or genuinely insufficient coil capacity? Fix airflow distribution and containment before changing setpoints. When evaluating a setpoint change, consider the effect on coil leaving temperature, chiller lift, humidity control and fan energy together, and check against the equipment's allowable inlet envelope.
Overlooking humidity control interactions: assuming humidification is always needed, or that economiser and free-cooling modes do not affect humidity, and ignoring the sensible versus latent load split.
Humidity is taught as a fixed band, so candidates apply humidification and dehumidification as routine. In modern halls with wide allowable envelopes and sealed containment, the latent load is small, and economiser operation can introduce or remove moisture depending on climate and mode.
What to do instead
Treat humidity as a control loop with a real latent load: identify the source of moisture (outdoor air, infiltration, people) and the removal path (coil condensation, dedicated dehumidification). Recognise that in a well-sealed hall with little outdoor air, simultaneous humidification and dehumidification is usually a control fault, not a design requirement, and check economiser mode against the allowable envelope.
Reading a psychrometric or airflow question without tracking units, mixing volumetric airflow (m3/s or cfm) with mass flow, and using total pressure where static pressure is meant.
Airflow and capacity calculations chain several unit conversions, and candidates carry cfm into a formula expecting m3/s or forget air density corrections at altitude and temperature. Fan and duct questions also hinge on static versus total pressure, which are often used loosely.
What to do instead
Write the unit next to every number and convert before substituting. Use mass flow (kg/s) for energy balances and volumetric flow only where the question asks for it, correcting for density where altitude or temperature matters. For fans and ducts, state whether the question concerns static, velocity or total pressure and use the matching curve.
Answering maintenance and efficiency questions with generic advice ('clean filters, do preventive maintenance') instead of the specific measure and its measurable effect, such as coil cleaning restoring approach temperature or economiser hours reducing compressor run time.
Operations content is easy to answer in generalities, and candidates do not connect a maintenance action to the metric it moves. This loses marks on questions that ask which action gives the largest efficiency gain or how to verify it.
What to do instead
Pair every operational action with its indicator: filter change to static pressure drop and fan power, coil cleaning to approach temperature and capacity, belt and bearing checks to airflow, valve and damper calibration to control stability, and economiser verification to hours of mechanical cooling avoided. When asked for the best measure, rank by measured impact on PUE or capacity, not by how often it is scheduled.